Analysis of criteria for grading bladder cancer in urine cytological tumor diagnosis by means of an expert system.

Abstract

An inductive expert system was used for the analysis of criteria for grading bladder carcinoma in urine cytological tumor diagnosis. This analysis seems necessary in order to provide a better standardization of grading and to avoid tumor grades, which are rather inhomogeneous with respect to morphology and prognosis. The analysis of the database by the inductive system shows a considerable variation of the cytomorphology of different bladder carcinomas graded as G2 tumors, whereas G1 and G3 tumors are more homogeneous groups respectively. Especially nuclear morphological criteria are important features for the detection of highly differentiated carcinomas, whereas nucleolar features might be helpful to assess the proliferative nature of the carcinoma. The future goal of avoiding a grading system with prognostically inhomogeneous tumor grades seems possible when using an inductive expert system for consultation.